The MHT building dataset was collected for purposes of testing RGB-D mapping of changing environments. The dataset consists of 2D (laser scanner) and 3D (Asus Xtion) data collected by a SCITOS-G5 mobile robot. We provide the gathered data in form of rosbags.
The intended use was to create a dataset to benchmark spatio-temporal representations of changing environments. The robot was patrolling a small office (see the video below) every 5 minutes. Each patrol started and ended at a charging station. During each patrol, the robot continuously collected its laser scans and odometric data. Moreover, it stopped at three different locations, took snapshots using its RGB-D sensor and attempted to detect people presence.
| MHT office night collections with 3D sweeps |
The provided archives contain rosbags, which are zipped into separate files according to the day of the data collection and data type. Each rosbag with a 3D prefix contains a depth/color image, camera information, robot position, tf data, laser scan and person detection gathered by the robot at a location and time that is encoded in the rosbag name, which contains day, month, year, hour, minute and location id. For example, 3D_23-08-13-15-20_place_2.bag contains data gathered at location 2 on August 23 2013 at 15:20 o'clock. Each rosbag with a 2D prefix contains AMCL position estimates, robot odometry, tf data and laser scans. Day, month, year, hour and minute are part of the bag file name.
If you use the dataset for your research, please cite our paper that describes the data collection in detail. We attached a bibtex record for your convenience.